DeepRL Parking
Project Objective
The objective is to design and implement a goal-conditioned continuous control task for an ego-vehicle, requiring it to autonomously navigate and park within a predefined space while aligning with the specified heading.The objective is to design and implement a goal-conditioned continuous control task for an ego-vehicle, requiring it to autonomously navigate and park within a predefined space while aligning with the specified heading.
Contributors
Abstract
Autonomous parking systems have gained significant attention in recent years for their potential to improve driving safety, convenience, and efficiency. Deep Reinforcement Learning (Deep RL) offers a promising approach for training agents to navigate complex environments and make optimal decisions in real time. This report reviews the application of Deep RL techniques to autonomous parking, starting with the key challenges such as continuous state and action spaces, and decision-making under uncertainty. It provides an overview of Deep RL and its capability to learn complex behaviors through interaction with the environment and reward feedback. The report also discusses various Deep RL architectures—such as Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), and Soft Actor-Critic (SAC)—in the context of autonomous parking.
Report
Presentation
⬇️ Download Full Presentation (PDF)